← Agent Seal · Register · Cases · Research · Blog · Compliance

EU AI Act readiness for autonomous agents

How the Agent Seal agent registry maps to obligations under Regulation (EU) 2024/1689, and where our partner aisthetix.de covers organisation-level assessment and data governance.

This is a readiness aid, not legal advice or a compliance verdict. Article numbers, Annex references, and application/registration dates are subject to change. Nothing here should be published as a compliance claim without review by qualified counsel. "Supports" and "helps satisfy" never mean "makes you compliant".

Use the registration guide to prepare a new agent record, then use the public verifier to check a live agent's signed Agent Card, registry status, EU evidence bundle, court export, append-only event history, and recorded PEDIGREE intent chain from one browser surface. For plain deployment patterns, see the pilot case studies.

Near-term timing: Article 50 transparency applies from 2 August 2026. The Commission's current AI Omnibus framing moves standalone Annex III high-risk systems toward 2 December 2027 and product-integrated high-risk systems toward 2 August 2028. The high-risk classification consultation closes 23 June 2026, 22:00 CET.

The partnership

The AI Act is not sector-based — it applies by role (provider vs deployer, Art. 3 / 25) and the risk tier of the specific use (Art. 5 / 6 / 50). Two layers, two partners:

aisthetix.de — organisation readiness

Role + risk-tier assessment, prohibited-use screen (Art. 5), high-risk deployer/provider readiness (Art. 6, 26/27), transparency (Art. 50), GPAI (Art. 53/55), and DSGVO/GoBD data governance (Art. 10) — as a scored, sovereign readiness self-assessment.

aisthetix.de →

agent-seal.xyz — agent accountability

Per-agent technical controls and evidence: verifiable identity, a signed covenant, a human liability anchor, append-only traceability, a signed Agent Card, JWKS, and a court-grade export.

agent-seal.xyz →

What agent-seal.xyz provides, by article

These are technical building blocks that contribute evidence toward each obligation — they do not discharge it on their own.

ProvisionObligationSupporting feature
Art. 12Record-keeping / automatic logging over the lifetimeAppend-only Merkle-DAG trace log, hash-chained, OpenTimestamps-anchored, court-export package.
Art. 13 & 50Transparency / disclosure of AI interactionPublic signed Agent Card, JWKS, seal + /v1/seal/verify, public status endpoint.
Art. 14Human oversight; ability to interveneLiability anchor (agent tethered to a human keeper); standing supports suspend / revoke.
Art. 17Quality management evidence hooksEvidence bundle, status, trace export, and standing changes that feed a provider QMS.
Art. 4AI literacy of operatorsHuman-readable covenant + certificate documenting scope, capabilities, keeper.
Art. 26Deployer obligations (logs, oversight, monitoring)Per-agent traces, keeper tether, live standing.
Art. 49 / 71Registration and EU database workflow/v1/public/eu-database prototype feed. Complements, does not replace, the official EU database.
Art. 73 / 85Incident and complaint referencesPublic procedure links and governance references without exposing private reports.
Art. 25 / 72Value chain & post-market monitoringArchitect→agent link, persisted standing changes, append-only log substrate.

Machine-readable evidence

API guide/api
Registration schema/agent-registration.json
Profile/.well-known/agent-accountability-profile.json
Signed card/.well-known/agent-card.json
JWKS/.well-known/jwks.json
Status/v1/public/agents/{id}/status
Court export/v1/public/agents/{id}/court-export
Redaction policy/v1/public/evidence-redaction-policy
Evidence/v1/public/agents/{id}/eu-ai-act-evidence
Report/v1/public/agents/{id}/eu-compliance-report
EU database feed/v1/public/eu-database
GPAI/v1/public/gpai-models
Templates/v1/public/compliance/templates

Check an agent's readiness

Agent-level technical controls only — not an organisation-level verdict.

References: Regulation (EU) 2024/1689, Arts. 3–14, 25–27, 49–55, 72; Annex III. All mappings require validation by qualified counsel before any public compliance claim.